TY - JOUR
T1 - UCDFormer
T2 - Unsupervised Change Detection Using a Transformer-Driven Image Translation
AU - Xu, Qingsong
AU - Shi, Yilei
AU - Guo, Jianhua
AU - Ouyang, Chaojun
AU - Zhu, Xiao Xiang
N1 - Publisher Copyright:
© 1980-2012 IEEE.
PY - 2023
Y1 - 2023
N2 - Change detection (CD) by comparing two bitemporal images is a crucial task in remote sensing. With the advantages of requiring no cumbersome labeled change information, unsupervised CD has attracted extensive attention in the community. However, existing unsupervised CD approaches rarely consider the seasonal and style differences incurred by the illumination and atmospheric conditions in multitemporal images. To this end, we propose a CD with a domain shift setting for remote sensing images. Furthermore, we present a novel unsupervised CD method using a lightweight transformer, called UCDFormer. Specifically, a transformer-driven image translation composed of a lightweight transformer and a domain-specific affinity weight is first proposed to mitigate domain shift between two images with real-time efficiency. After image translation, we can generate the difference map between the translated before-event image and the original after-event image. Then, a novel reliable pixel extraction module is proposed to select significantly changed/unchanged pixel positions by fusing the pseudochange maps of fuzzy c-means clustering and adaptive threshold. Finally, a binary change map is obtained based on these selected pixel pairs and a binary classifier. Experimental results on different unsupervised CD tasks with seasonal and style changes demonstrate the effectiveness of the proposed UCDFormer. For example, compared with several other related methods, UCDFormer improves performance on the Kappa coefficient by more than 12%. In addition, UCDFormer achieves excellent performance for earthquake-induced landslide detection when considering large-scale applications. The code is available at https://github.com/zhu-xlab/UCDFormer.
AB - Change detection (CD) by comparing two bitemporal images is a crucial task in remote sensing. With the advantages of requiring no cumbersome labeled change information, unsupervised CD has attracted extensive attention in the community. However, existing unsupervised CD approaches rarely consider the seasonal and style differences incurred by the illumination and atmospheric conditions in multitemporal images. To this end, we propose a CD with a domain shift setting for remote sensing images. Furthermore, we present a novel unsupervised CD method using a lightweight transformer, called UCDFormer. Specifically, a transformer-driven image translation composed of a lightweight transformer and a domain-specific affinity weight is first proposed to mitigate domain shift between two images with real-time efficiency. After image translation, we can generate the difference map between the translated before-event image and the original after-event image. Then, a novel reliable pixel extraction module is proposed to select significantly changed/unchanged pixel positions by fusing the pseudochange maps of fuzzy c-means clustering and adaptive threshold. Finally, a binary change map is obtained based on these selected pixel pairs and a binary classifier. Experimental results on different unsupervised CD tasks with seasonal and style changes demonstrate the effectiveness of the proposed UCDFormer. For example, compared with several other related methods, UCDFormer improves performance on the Kappa coefficient by more than 12%. In addition, UCDFormer achieves excellent performance for earthquake-induced landslide detection when considering large-scale applications. The code is available at https://github.com/zhu-xlab/UCDFormer.
KW - Change detection (CD)
KW - UCDFormer
KW - domain shift
KW - transformer
KW - unsupervised learning
UR - http://www.scopus.com/inward/record.url?scp=85171534450&partnerID=8YFLogxK
U2 - 10.1109/TGRS.2023.3305334
DO - 10.1109/TGRS.2023.3305334
M3 - Article
AN - SCOPUS:85171534450
SN - 0196-2892
VL - 61
JO - IEEE Transactions on Geoscience and Remote Sensing
JF - IEEE Transactions on Geoscience and Remote Sensing
M1 - 5619917
ER -